You do not need to know who will buy the product you imagine. You need to find out who experiences the problem, in what circumstances, and which differences matter when studying it.

Early segmentation: reduce vagueness without closing off the market

Saying “my customer is anyone who wants to eat better” tells you little about whom to interview, where to observe behavior, or which alternatives to compare. Initial segmentation identifies groups that help you understand the problem: employees who buy lunch near the office, students with changing schedules, and families planning weekly meals. It does not guarantee that any of them will buy, nor does it force you to choose a permanent market.

The SBA considers consumer behavior, economic data, location, and demand relevant dimensions of market research [R01]. GOV.UK's guidance reminds us that, at the start of discovery, we may not even know who the users will be: identifying them is part of the research [R03].

Four dimensions that can help you segment

Situation and task. What is the person trying to accomplish, and when? A restaurant that handles many weekend bookings may have different needs from one that serves walk-in customers.

Current behavior. What do they do today, how often, through which channel, and why? When exploring an opportunity, observed behavior is often more useful than creating groups based only on age or income.

Constraints and resources. Time, budget, equipment, location, skills, regulations, or family responsibilities. These attributes matter when they explain how the problem is experienced.

Relationship to the purchase decision. Do they use, recommend, decide, approve, or pay? In B2B settings these roles may belong to different people. Do not interview only end users if adoption requires approval from an owner or purchasing department.

You do not need every possible variable. Choose variables that could change how you research or later design a response. Segments that you cannot distinguish or contact are not very useful at this stage.

Potential market, target market, and first interviewees

These are different levels. The potential market is the broad universe of people or organizations that might share relevant conditions. The preliminary target market is the subset you choose to focus your exploration on because of an apparent need and accessibility. Research participants are actual individuals selected to understand experiences within—and, where useful, outside—that subset.

For example, a region may have many restaurants, but researching reservations requires establishments that actually take them. Within that group, you may start with businesses receiving bookings through several channels and employing few people. Do not automatically turn the regional number of restaurants into the number of potential buyers; that distinction belongs in the demand estimate.

To avoid confusing an interesting segment with the entire available market, see how to estimate market size and demand.

How to build initial segmentation step by step

1. List situations, not imaginary archetypes: booking a table, changing an appointment, repeating a purchase, fixing a breakdown, restocking a supply. 2. Identify actors and who bears the consequences. 3. Select observable variables that may explain differences: volume, frequency, channel, schedules, decision-making autonomy. 4. Build a matrix of two to four provisional segments. 5. Recruit people using verifiable criteria, including cases where the problem is absent. 6. Revise the groups when patterns contradict your first categories.

Exploratory segment Situation What we need to learn
Restaurants receiving bookings through several channels Changes and confirmations Where information is lost and how mistakes are avoided
Restaurants using a notebook with low demand Simple management Why their existing alternative works
Establishments not taking reservations Walk-in service Whether the assumed problem is irrelevant

This table is illustrative. It does not imply these segments occur in any particular proportions in a real market.

Once you know which groups you need to understand, prepare discovery interviews with relevant participants.

What a buyer persona is—and is not

A buyer persona usually summarizes a type of customer or purchaser with relevant needs, motivations, contexts, and behaviors. In UX, the term user persona is common; a user persona does not necessarily describe the economic buyer. Nielsen Norman Group distinguishes personas based on qualitative research, statistical personas, and proto-personas, which organize assumptions without pretending those assumptions have been established [R13].

A persona is not merely “Martha, 32, likes coffee,” unless those details help you make a specific decision. For researching reservation services, it matters much more whether Martha manages bookings during service, how much time she has, and how she deals with disruptions. Her age and gender may explain nothing about that work. Inventing characters with names, photos, and hobbies does not transform a description into knowledge.

How to build a persona using available evidence

Once you have observed patterns, define the purpose: “We need to represent the person who manages bookings during service.” Gather evidence from interviews and observations, grouping recurring goals, behaviors, and barriers. Create a profile covering role or context, primary task, use situation, current alternatives, limitations, decision criteria, and traceable quotes or incidents. State which findings support it and what remains hypothetical.

If no research has taken place, label the result a proto-persona and use it to plan what to check. If two groups share most of their needs, one profile may work better; if tasks and decisions differ, separate profiles may be useful. NN/g notes that a persona's detail and scope should fit its intended use [R13, R14].

Example: a natural products store

Imagine you want to improve the experience of repeat purchasing. One segment might consist of customers who take supplements daily and buy another container every few weeks; another might consist of occasional customers buying products as gifts. Apparent frequency is a lead, not a proven regularity: people may pause consumption or vary doses and package sizes.

You could start by studying repeat purchasers, observe how they remember to restock a product, and ask what they did the last time. A proto-persona saying “wants automated reminders” would introduce a solution without evidence. A research-based profile might say “in several observed purchases, the customer could not remember when the container would run out,” while retaining the number and context of those cases.

Mistakes and limitations

Confusing personas with statistical segments. A narrative profile does not estimate how many buyers exist. Segmenting by every available variable. A variable matters only if it explains a decision or context. Choosing only enthusiastic users. It is also useful to learn why some people do not have the problem. Forgetting the buyer. The person using the solution may not be the one paying for it. Treating categories as permanent. Research segmentation should change when better evidence appears.

What you should have when you finish

A matrix of provisional segments, recruitment criteria, identified users/buyers/decision-makers, outstanding questions for each group, and—only if useful—one or two persona profiles supported by research or clearly labeled as proto-personas. The final commercial segment and channel choices belong to business design; here we are building a defensible understanding of people.

For more detail, see how to identify, organize, and prioritize customer needs and how to design the customer journey and relationships.

Main sources: [R01] SBA; [R03] GOV.UK; [R06] recruitment; [R13] NN/g persona types; [R14] NN/g scope. The example is hypothetical.

Bibliography and references